I am wondering if anyone has had success using the package ohun for energy-based detection and successful evaluation of their detection routines. I have used the vignette to learn a lot, but I am coming upon regular challenges and would love to speak with others about it.
My main challenge with ohun right now has to do with the detection of true positives calculated upon using optimize_energy_detector(). I noticed that the number of “Total Detections” and “True Positives” is always the same, when I thought (based on signal detection theory and the vignette) that the number of total detections should probably include the sum of true positives and false positives. Perhaps more importantly, I am finding that when I use optimize_energy_detector(), I find MORE true positives in the sound file than exists in the reference file I have created (selection table). For example, my reference table has 24 sounds called snap-jumps, and optimize_energy_detector() counts 59 total detections, 59 true positives, and 10 false positives. Somehow, the package is finding more sounds - and calling them true positives - than there could possibly be. This does not make sense to me, and does not make sense based on the description of how the package works (I could be missing something). I saw on this page that @lostanlen mentioned a problem with counting true positives, and I wonder if that is related to the problem I'm having. Thanks in advance folks.
Kristen
